H-AGI = Human led Artificial General Intelligence
Concept: Hybrid Intelligence
The Objective: Using AI as a tool so that it augments, rather than diminishes, your IQ over time.
Warning: Extensive, uncritical gen AI use has significant potential to reduce IQ over time.
H-AGI Use Cases (General)
Buy the right product (phone, laptop, etc.)
Human IQ: Set goal & constraints (budget, size, battery need). Write IF→THEN. Define pass/fail (e.g., “≥8h in 2 independent tests”). Decide.
AI: Librarian (3 only) to fetch test-based reviews; extract quotes; Critic to propose the quickest refuter (e.g., stress test that model vs a rival).
Plan a short trip/weekend
Human IQ: Set constraints (≤£250, <2h travel, accessibility). Draft 1-page plan. Choose mode (Compress/Expand). Decide trade-offs.
AI: Compress to a minimal itinerary that meets constraints; or Expand one element (e.g., swap lodging). Produce checklist to verify constraints.
Start a health habit micro-pilot (non-medical)
Human IQ: Map mechanism; IF→THEN; pick one metric (e.g., restlessness 0–10). Set pass/fail (≥4/7 days show ≥2-point drop). Log results & decide.
AI: Expand one element (timing/duration), draft a 7-day tracker row, and Critic to suggest the smallest decisive check.
Tame spending with a mini-budget
Human IQ: Choose one category + one rule (e.g., £60/wk eating out). Define “no spillover” constraint. Decide after 2 weeks.
AI: Compress the plan to the few behaviours that change outcomes; generate a 2-week tracker and an invariances list (what must stay fixed).
Improve CV/cover letter & interview hit-rate
Human IQ: Identify role target; write IF→THEN (tailoring → more callbacks). Run A/B across 10 apps. Decide next step.
AI: Expand “tailoring” once (JD keywords + proof bullets); Critic to define the A/B threshold (e.g., ≥20% higher response) and a pass/fail table.
Quick news/claim check (misinfo hygiene)
Human IQ: Predict first (what figures you expect). Decide credibility rule (e.g., 2 convergent reputable sources).
AI: Librarian (quotes) to return ≤3 sources with ≤25-word quotes; Critic to name a rival interpretation and a fast refuter.
H-AGI Use Cases (Students)
Essay/argument shaping (coursework or blog)
Human IQ: Draft thesis + IF→THEN; list rival. Decide which claims to keep/drop.
AI: Librarian (3 only) for directly relevant sources; Critic to stress-test with one decisive comparison; Compress to a tighter thesis.
Research proposal scoping & hypothesis refinement
Human IQ: Map mechanism; write IF→THEN; set primary DV and pass/fail. Decide final design.
AI: Compress to prediction-changing parts; outline a 1-day pilot with thresholds; format aims/hypotheses cleanly.
Literature review & alignment (with PDFs/notes)
Human IQ: Define inclusion rule (e.g., “Keep if ≥2 support and none contradict”). Make the keep/drop call.
AI: Consistency Gate (attachments only) to build Supports | Contradicts | Silent with quotes + page/time; generate a 5-item Fix List.
Coaching / habit change micro-pilot
Human IQ: Choose one element to test (e.g., 5-min breathing), write predictions + must-hold, set pass/fail, run the week, decide next step.
AI: Expand to place the element on your map; propose the smallest test; Reframe Log to assemble Keep/Drop/Open and the next smallest action.
Quick reminder
Human IQ = ownership: goal, constraints, mental map, IF→THEN, pass/fail rule, decisions.
AI = support roles: Librarian, Critic, Compress/Expand, Consistency Gate, and clean logging.
One prompt per beat and always log Sources & Changes at the end.
Here’s a simple checklist you can use as an entry-level H-AGI playbook.
The 5-Beat Loop (run in order)
Open every loop-pass with a one-liner goal and 2–3 red lines (constraints).
(Constraints = hard limits you won’t break, e.g., “≤ 30 minutes”, “must follow APA 7”, “no new data collection”, “≤ 500 words”.)
You need a Notes document and the AI chat window. (A simple spreadsheet can help track predictions and results.)
1) Diverge (you first — predict before you look)
Your own cognitive map (3–5 boxes with arrows) in your own words.
e.g., “Stress → poor sleep → lower focus → lower grades.”
(Pen & paper is fine—scan a sketch if needed.)One IF→THEN (mechanism).
e.g., “IF a 5-min breath drill lowers arousal, THEN same-session stress should drop.”Add 1–2 predictions (metric / timing / comparison).
e.g., “On drill days, pre→post stress drops ≥ 2 points; evening focus is above my 7-day baseline (baseline = your pre-start 7-day average).”State one must-hold (invariance).
e.g., “Effect is immediate (same session), not next day.”Set a simple pass/fail rule.
e.g., “Promote if ≥ 4/7 days meet the ≥ 2-point drop; otherwise sandbox/adjust.”(Optional) Self-test: ask the AI to make 6 questions from your notes only; answer, then let it mark ✓/✗ with one-line fixes.
(Optional — one button only) Librarian (3 only): ask for ≤ 3 sources that directly test your IF→THEN or P1/P2 (no new concepts yet).
2) Gate (10-second mode check)
Type/decide: “Foggy or clear?”
Foggy → Expand (Creative)
Clear → Compress (Control)
3a) Compress (Control) — if Clear
Rewrite your idea in 1–2 sentences using your words.
List only the parts that change predictions; make a Drop list for the rest.
Prompt yourself: “Does this part change what I’d expect to happen?” (✓/✗)(Optional) Run a Self-test.
3b) Expand (Creative) — if Foggy
Add exactly one new piece X (one idea, tool, or link).
Example X: “5-minute 6-breaths/min drill before study.”Show where X plugs into your map (list arrows only).
Write 2 predictions + 1 must-hold tied to X.
Must-hold = a condition that must be true for X to belong (e.g., “effect is short and immediate, not next day”).Propose the smallest decisive check you could do today.
e.g., “Track pre→post stress for 7 days; pass if ≥ 4/7 show ≥ 2-point drop.”(Optional) Run a Self-test.
4) Stress-test (try to break it as AI critic)
Name one rival explanation.
Choose the quickest check with a simple pass/fail rule in plain English.
(Optional; attachments only) Consistency Gate (attachments only): build a Supports | Contradicts | Silent table with short quotes + page/time, then write a 5-item Fix List.
5) Reframe (learn & move)
Make three lists: Keep / Drop / Open questions.
Write the smallest next step that will teach you the most.
Log “Sources & Changes” (what changed in your thinking and why).
Guardrails (always on)
You think first. Draft in your own words before asking the AI.
No new concepts from AI unless you open the Expand Gate.
One piece at a time in Expand (add one element, then test).
Smallest decisive check beats big, slow plans.
Leave a trail every pass (update Sources & Changes).
One prompt per beat.
Quick Prompts to paste (one per beat)
Compress Pass (Control):
Rewrite my idea in 1–2 sentences (use my words), mark which parts change predictions (✓/✗), and propose a Drop list. No new terms.Open Expand Gate (Creative):
Open Expand Gate: X = <one new element>. Using only my map & commitments: place X (list arrows), write 2 predictions + 1 must-hold, propose one smallest test (today). No new primitives beyond X.Critic (decisive check):
Name one rival, propose the quickest refuter, and give a plain pass/fail threshold. No new parts.Librarian (quotes — web/general):
Return ≤3 sources that directly test my IF→THEN or P1/P2. For each: 1-line why it matters; one short quote (≤25 words) with citation; tag Supports/Contradicts/Mixed; study type + key comparison. No new concepts; prefer recent primary/meta-analytic evidence.Librarian (quotes — attachments only):
From my attachments only, return up to 5 short quotes (≤25 words) that directly test my IF→THEN or P1/P2. For each: doc name + page/time, 1-line mapping, tag Supports/Contradicts/Silent. No new ideas.Reframe Log:
List Keep / Drop / Open questions, propose the smallest next step, and append to my Sources & Changes (bullets).
Stop rule: When your 20–30 minutes are up, do the next smallest action you wrote. Start the next pass later with Goal + constraints again.
That’s it — one mental map, one IF→THEN, pick your mode, make one small check, and leave a trail.
Try it!
Let me know if this method (or any element from it) works in attaining your goal, while ensuring you are in charge of the cognitive process.
What is critical with the increasing use of IQ is to ensure we maintain self-efficacy with our own human IQ!